Data processing: generic control systems or specific application – Specific application – apparatus or process – Product assembly or manufacturing
Reexamination Certificate
2000-02-01
2003-03-25
Picard, Leo (Department: 2125)
Data processing: generic control systems or specific application
Specific application, apparatus or process
Product assembly or manufacturing
C702S084000
Reexamination Certificate
active
06539272
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to an inspection method and system to specify manufacturing processes in which measures should be taken against particles or pattern defects in the manufacture of a wafer. In this specification, particles and/or pattern defects are simply called defects.
Yield is a good dies ratio which is calculated using the results of a probing test, which is final electrical test in the process of manufacture of a wafer. That is, yield is a ratio of good dies to all dies of a wafer. To improve yield, it's important to calculate the yield impact of each manufacturing process and to take measures in the manufacturing process which has great yield impact.
Prior methods to quantify yield impact are as follows:
(1) A Correlation Analysis Between Yield and the Number of Defects
In this method, particle inspection or pattern defect inspection is performed every time a layer is formed on a wafer, and then the correlation analysis between yield and the number of defects is calculated based on results of a particles inspection or pattern defect inspection (See FIGS.
4
(
a
) and
4
(
b
)). According to this method, users can determine yield impact statistically.
However, this method must be performed on the premise that defects occur randomly on the wafer and that the inspection results follow the Poisson distribution. As shown in FIG.
4
(
a
), the situation wherein yield decreases as the number of defects increases is an ideal correlation relation between yield and the number of defects. The slope of correlation relation of FIG.
4
(
a
) represents yield impact per single detect. On the other hand, as shown in FIG.
4
(
b
), in the case wherein clustered defects occur or defects concentrate at one area in a wafer due to equipment problems, the correlation relation between yield and the number of defects becomes indistinct, or in the worst case, the yield is likely to increase as the number of defects increases. This situation is opposite that of the ideal correlation relation shown in FIG.
4
(
a
).
Methods for solving these problems are disclosed in the papers “Modification of Poisson statistics: Modeling defects induced by diffusion” written by O. Paz et al in the IEEE Journal of solid-state Circuits, vol. SC-12, pp. 540-546 (1977), and “Clustered Defects in IC Fabrication: Impact on Process Control Charts” written by D. J. Friedman et al in the IEEE Transactions on Semiconductor Manufacturing, vol. 4, No. 1 (1991). However, these prior methods can't take measures against false detection by an inspection tool.
And, in the paper “Statistical Micro Yield Modeling” written by Allan Y. Wong, in Semiconductor International, November, 1996, pp. 139-148, yield elements are resolved into systematic elements and random elements, and then the correlation relation of each element, between yield and the number of defects is analyzed. However, this prior method also can't take measures against false detection by an inspection tool.
(2) Kill Ratio Analysis Method
This method is disclosed in the paper “Yield Monitoring and Analysis in semiconductor manufacturing” written by S. Hall et al in the SEMICON KANSAI '97, ULSI technical seminar, pp. 4/42-4/47, (1997). As shown in
FIG. 5
, particle inspection or pattern defect inspection are performed every time each layer is formed on a wafer, and then the number of good dies a, c and the number of bad dies b, d are calculated based on the result of the final test(probing test). The value “a” is the number of good dies in the dies without particles. The value “b” is the number of bad dies in the dies without particles. The value “c” is the number of good dies in the dies with particles. The value “d” is the number of bad dies in the dies with particles. After this calculation, the yield of the dies without defects and the yield of the dies with defects are calculated, and the kill ratio of each layer is also calculated. According to the kill ratio obtained by this method, users can determine the number of bad dies which contain by defects. And, according to this method, the yield impact can be calculated without influence from process margin failure, such as process size and film thickness. However, this prior method also can't take measures against false detection by an inspection tool.
(3) Critical Area Analysis Method
This is a method to determine kill ratio depending on the area in the die by checking the a CAD data of circuit pattern with the size or position of defects. The method of determining kill ratio is disclosed in the paper “Modeling of defects in integrated circuit photolithographic patterns” written by C. H. Stapper in the IBM Journal of Research and Development, Vol. 28, No. 4, July, 1984, pp. 461-475, and in the paper “Modeling the Critical Area in Yield forecasts” written by A. V. FerrisPrabhu in the IEEE Journal of Solid-state Circuits, Vol. SC-20, No. 4, August, 1985, pp. 874-877. U.S. Pat. No. 5,598,341 discloses a yield control system using a critical area analysis method. This method can analyze more accurately than both the correlation analysis method and the kill ratio analysis method because it uses a circuit pattern. However, this prior method also can't take measures against false detection by an inspection tool and clustered defects.
False detection by an inspection tool means that non-real particles or nonreal pattern defects are, detected by the inspection tools as real particles or real pattern defects. For example, in case a film condition on a wafer is different from a normal film condition, if the sensitivity of the inspection tools is high, the film condition is likely to be detected as real particles or real pattern defects. And, if the lighting of the inspection tool changes value, the inspection tool is likely to detect non-real particles or non-real pattern defects as real particles or real pattern defects.
As mentioned above, the prior art can't take measures against false detection by inspection tools.
SUMMARY OF THE INVENTION
As mentioned above, the prior art can't take measures against false detection by inspection tools. Therefore, it is difficult to quantify yield impact acolirately by using the prior yield impact quantification methods because the prior methods are influenced by clustered defects, peculiar detect distribution on a wafer, false detection by inspection tools, and a disturbance such as process margin failure.
And, the prior methods require extraction of adder defects. As shown in
FIG. 6
, adder defects are defects which have occurred on a wafer newly. For example, in
FIG. 6
, detected defects which are detected at first inspection (1) are adder defects of inspection (1). Next, at inspection (2), the detected defects which are detected at inspection (1) are taken from detected defects which are detected at inspection (2), and then the result of this calculation becomes adder defects of inspection (2). Next, at inspection (3), detected defects which are detected at inspection (1) and detected defects which are detected at inspection (2) are taken from detected defects which are detected at inspection (3), and then the result of this calculation becomes adder defects of inspection (3). For the remainder of the inspection (4~), the same calculation is performed to calculate adder defects of each inspection step.
However, extraction of adder defects is influenced by reproductivity of the X-Y stage of the inspection tool or alignment error. And, cluster defects are likely to be detected differently at each of the inspection steps; especially, the size of the cluster defects will be detected differently at each of the inspection steps. Therefore, it is difficult to check one layer's defects with another layer's defects correctly.
The purpose of the invention is to quantify yield impact accurately, especially to quantify yield impact accurately without extraction of adder defects.
To achieve the purpose of the invention, the manufacturing line is controlled in accordance with the amount of yield impact of dies w
Iwata Hisafumi
Nemoto Kazunori
Ono Makoto
Antonelli Terry Stout & Kraus LLP
Cabrera Zoila
Hitachi , Ltd.
Picard Leo
LandOfFree
Electric device inspection method and electric device... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Electric device inspection method and electric device..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Electric device inspection method and electric device... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3069930